Transfection of AtT-20ins Cells with GLUT-2 but Not GLUT-1 Confers Glucose-stimulated Insulin Secretion RELATIONSHIP TO GLUCOSE METABOLISM*

Glucose is thought to stimulate insulin release from islet &cells through generation of metabolic signals. In the current study we have introduced the genes encoding the facilitated glucose transporters known as GLUT-1 and GLUT-2 into AtT-ZOins cells to assess their impact on glucose-stimulated insulin release and glucose metabolism. We find that transfection of AtT-20in. cells with GLUT-2, but not GLUT-1, confers glu- cose-stimulated insulin release in both static incubation and perifusion studies. Cells transfected with GLUT-1 have a K, for 3-0-methyl glucose uptake of 4 mM and a v,, of 5-6 mmol/min/liter Cell space. These values are increased compared to untransfected AtT-20i, cells (K, = 2 mM; V,, = 0.5 mmol/min/liter cell space), but are less than observed in GLUT-2-transfectedlines (K, = 16-17 mM; Vmax = 17-25 mmoll min/liter cell space). Despite these dramatic differ- ences in glucose transport affinity and capacity, the rates of [5-SH]glucose usage are not different in the control and transfected lines over a range of glucose concentrations from 10 pM to 20 mM. We conclude that the specific effect of GLUT-2

Glucose metabolism appears to be required for glucosestimulated insulin release from pancreatic islet &cells. The control of glucose metabolism in P-cells is thought to reside mainly at the level of glucokinase-catalyzed phosphorylation of glucose (reviewed in Ref. l ) , but an important permissive role for the low affinity facilitated glucose transporter known as GLUT-2 has also been suggested (2)(3)(4)(5). In recent studies, we have been investigating the utility of the non-islet cell line AtT-ZOi,. for studies of the specific roles of GLUT-2 and glucokinase in the control of glucose-stimulated insulin release (reviewed in Ref. 5). These cells are derived from corti-* This work was supported by National Institutes of Health Grant Pol-DK42582 (to C. B. N. and J. H. J.) and a grant from the Diabetes Research and Education Foundation (to C . B. N.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
7 To whom correspondence should be addressed Gifford Laboratories for Diabetes Research & Dept. of Biochemistry, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75235. Tel.: 214-688-2930;Fax: 214-688-8291. cotropin-secreting cells of the anterior pituitary and have been engineered for constitutive expression of human insulin (6). We established that the islet isoform of glucokinase is naturally expressed in AtT-20in. cells and were able to detect low levels of its enzymatic activity (20-30% of the activity in normal islets), but also showed that the cells lack natural expression of GLUT-2 and fail to respond to glucose (7). Upon stable transfection with GLUT-2, AtT-ZOi,, cells exhibit increased insulin content, glucose potentiation of non-glucose secretagogues, and a direct stimulation of insulin release by glucose, albeit with maximal effect at subphysiological concentrations of the sugar (8). We concluded that GLUT-2 expression in this endocrine cell line allowed glucose-regulated hormone release and that the response to subphysiological levels of glucose (maximal at 10-50 pM) was consistent with the relative predominance of low K,,, hexokinase activity in these cells.
The current work was directed at determining whether glucose-stimulated insulin secretion can be conferred in AtT-20i,, cells by overexpression of GLUT-1, an alternate member of the family of facilitated glucose transporters. Glucosestimulated insulin secretion was measured in multiple GLUT-1-and GLUT-%transfected AtT-BOi,, cell lines both by static incubation and perifusion techniques, thereby allowing a detailed evaluation of the magnitude and dynamics of any glucose-induced responses. By measuring 5-[3H]glucose usage in GLUT-1uersus GLUT-2-transfected lines, we have also attempted to dissect the relative importance of glycolytic flux uersus transporter isoform expression per se in mediating glucose sensing.

Stable Transfection of AtT-BOi, Cells with GLUT-1-A 1.8-kilobase
Sal1 fragment of the cDNA encoding human GLUT-1 (a gift from Dr. Graeme Bell, University of Chicago) was subcloned into the vector PCB-7 immediately downstream of its cytomegalovirus promoter. The PCB-7 vector also contains a hygromycin resistance marker. The cloning process removed 660 bp' from the 3'-untranslated region and 84 bp from the 5"untranslated region of the GLUT-1 cDNA. . cells were transfected with this construct using electroporation as described previously (8). 15 stable transformants were isolated, grown for several passages in media containing 125 Kg/ml of hygromycin B (Boehringer Mannheim), and analyzed for GLUT-1 expression. As a control, cells were also selected with hygromycin following transfection with the PCB-7 vector lacking a GLUT insert.

15205
(IO), and as a control for RNA loading, a labeled 18 S rRNA oligonucleotide (9). Note, that although the GLUT-1 cDNA is of human origin, this should not significantly affect hybridization to the endogenous AtT-20i.. transcript, since the human and mouse GLUT-1 sequences are 97% identical (11,12). Immunofluorescent staining of GLUT-1 in AtT-20i, cells was carried out as previously described for GLUT-2 (8), using a 1:5000 dilution of commercial antibody preparation (East Acres Biologicals) directed at the rat GLUT-1 C-terminal peptide (100% identity with human GLUT-1 C terminus). The kinetics of 3-0-methyl glucose uptake was measured in GLUT-1-transfected AtT-aOi, cells as described previously (3,8).
Growth of AtT-20, Cell Lines in Liquid Culture and Evaluation of Secretagogue-induced Insulin Secretion-Cells were grown in liquid culture on gelatin microcarrier beads (Cultisphere-GL beads, Hy-Clone Laboratories) or Fibracel polyester discs (Bibby Sterilin Ltd.). Cultisphere beads were prepared for cell adhesion by washing with ethanol and coating with Matrigel (Collaborative Research Inc.) for 10 min at 25 "C; no coating of the polyester discs was required. Approximately 15 X lo6 cells were harvested from semiconfluent tissue culture plates and suspended in 25 ml of Dulbecco's modified Eagle's medium, supplemented with 10% fetal calf serum and 100 pg/ ml streptomycin/penicillin, and containing 100 mg (dry weight) of Matrigel-coated microcarriers or 3 g of Fibracel discs. The cell suspension was added to a 250-ml stirring flask, and cells were allowed to adhere to beads or discs for 18 h at 37 "C. Afterward, 100 ml of media was added, and the suspension was stirred continuously at 30 revolutions/min. Media was changed periodically, and cultures were typically grown for approximately 10 days prior to harvest of cells for perifusion or glucose usage experiments.
Insulin secretion from cells was evaluated by a modification of the column perifusion technique of Knudsen et al. (13). Cells were grown in liquid culture on the gelatin beads as described above, and 5-10 X 10' cells were harvested by gentle centrifugation (500 revolutions/ min in a Sorvall RT6000B desk top centrifuge) and resuspended in 10 ml of HEPES-bicarbonate buffered salt solution (HBSS), pH 7.4. The cell suspension was loaded into a Pharmacia C 10/10 column, and after the beads settled to the bottom, the top plunger of the column was inserted, and the whole apparatus submerged in a 37 "C water bath. Cells were washed for 15 min prior to the start of an experiment by perifusion with HBSS lacking glucose at a flow rate of 0.5 ml/min. Cells were then perifused alternately with HBSS lacking glucose or with HBSS containing 5 mM glucose for 25 min periods and with continuous collection of the effluent in 2.5 min fractions. A final 25-min perifusion with 5 mM glucose + 0.5 p M forskolin was performed at the end of all experiments.
Measurement of insulin secretion from cells grown in 12-well plates (static incubation assay) was performed as described previously (7). The insulin concentration in all samples was measured by radioimmunoassay and data normalized to protein content/well.
Measurement of Glucose Usage-Glucose usage was measured essentially by the method of Trus et al. (14). Cells grown in liquid culture on polyester discs as described above were placed into tubes containing multiple glucose concentrations over the range of 10 p M -20 mM. The discs, containing a total of between one and two million cells, were washed twice in HBSS containing the appropriate concentration of glucose, then equilibrated in 0.5 ml HBSS/glucose at 37 "C for 10 min. Measurements were begun by addition of 0.5 ml of HBSS containing [5-'H]glucose (Amersham Corp.) at various concentrations and with a final specific activity of 1-3 disintegrations/min/ pmol (approximately 1 mCi/mmol). Measurements were performed for 5 and 15 min and were terminated by addition of 250 pl of 10% perchloric acid. 100 pl of sample was then transferred to another tube, placed inside a capped scintillation vial containing 0.5 ml of water, and incubated at 50 "C for 18-24 h. After this vapor-phase equilibration step, the tube was removed from the vial, scintillation mixture (Bio-Safe; Research Products International) was added to the vial, and 3H20 content was determined by counting over a 5-min period. The efficiency of equilibration of the 'H20 was determined by aliquoting 1 pCi of 3H20 (in 100 pl) into tubes and measuring equilibration in these samples as for the experimental samples. The coefficient of equilibration (EQC) was defined by the fraction of total counts equilibrated from the tube to the vial for the 'H20 control after the incubation period. EQC varied between 0.6 and 0.7 in all experiments. Glucose usage was calculated as follows: (dpm sampledpm zero time) (specific activity (dpm/pmol glucose) X EQC X time (min) X protein (mg)

RESULTS
Transfection of AtT-20in. cells with the cytomegalovirus promoter/human GLUT-1 cDNA construct yielded 15 hygromycin-resistant clones. These were evaluated for GLUT-1 expression by blot hybridization analysis using a radiolabeled GLUT-1 antisense RNA probe. Fig. JA shows steady-state levels of GLUT-1 mRNA in five representative clones. Controls for the specific effects of GLUT-1 transfection included lanes containing RNA from the parental AtT-BOi,, cell line, the previously described GLUT-2 expressing cell line CGT-6 (8), and a rat insulinoma cell line RIN1046-38 (15). These three control cell lines all contained low to moderate levels of a 2.8-kilobase transcript corresponding to GLUT-1; the A. was prepared from five GLUT-1-transfected lines (designated I-2,l-5, 1-8, 2-10, and 1-12), from GLUT-2-transfected CGTB cells, rat insulinoma ( R I N ) 1046-38 cells, and untransfected AtT-2Oi.. cells (AtT). 10 pg of total RNA was loaded/lane, and blot hybridization was carried out sequentially with the antisense RNA or oligonucleotide probes indicated to the left of the panel, as described under "Materials and Methods." Panel B, total RNA was prepared from six GLUT-2-transfected lines (designated 24,2-8,2-9,2-12,2-14, and 2-1 7), from the previously described GLUT-2-transfected line CGT-6, untransfected AtT-20i.. cells (AtT) and from a GLUT-1-transfected line (1)(2)(3)(4)(5)(6)(7)(8). 20 pg of total RNA was loaded per lane, and blot hybridization was carried out sequentially with the antisense RNA or oligonucleotide probes indicated to the left of thepanel, as described under "Materials and Methods." Note, the increased 18 S rRNA signal in panel B relative to panel A is consistent with the fact that more RNA was loaded for this blot.
GLUT-1 probe also appeared to cross-react weakly with the highly expressed GLUT-2 transcript in the CGT-6 cell line. Four of the five GLUT-1-transfected lines (GT1-5,8,10, and 12) showed profound increases (50-100-fold) in levels of GLUT-1 mRNA, while a lesser increase (&fold) was observed in the fifth line studied (GT1-2). The GLUT-1 mRNA detected in the transfected cells is slightly smaller in size than that of native AtT-SOi,. cells due to partial deletion of the 3'untranslated region in the process of subcloning into the PCB-7 vector. In addition to the previously characterized GLUT-2-expressing AtT-20in, cell lines CGT-5 and CGT-6 (8), seven newly isolated GLUT-2 expressing clones were analyzed in this study. Five representative clones in Fig. 1B (GT2 -4,8,9, 12, 14) have GLUT-2 levels equal to or greater than the level in the CGT-6 cell line. For both panels of Fig. 1, 18 S rRNA probe hybridization is shown as a control for gel loading, and POMC mRNA levels were measured as a marker for the differentiated AtT-20 cell phenotype. All lines retained abundant levels of POMC mRNA. The blot in panel A was also probed with the labeled antisense insulin cRNA. Insulin mRNA was markedly elevated in CGT-6 cells relative to GLUT-1 transfected or parental AtT-20in, cells, consistent with our previous report of increased insulin content in GLUT-2 expressing AtT-20h. cells (8). Note that twice as much RNA was loaded in each lane for the blot of panel B compared to the blot in panel A. RIN1046-38 cells lacked POMC expression and had higher levels of insulin mRNA than any of the AtT-2Oi, lines. Two of the GLUT-1-expressing clones that contained high levels of GLUT-1 mRNA (GT1-10 and GT1-15) were used for immunocytochemical staining of GLUT-1 protein, using an antibody directed against the C terminus of GLUT-1. As shown in Fig. 2, line GT1-10 showed plasma membraneassociated GLUT-1 signal of much greater intensity than that seen in the parental cell line (data are shown for line GT1-10 only; expression in line GT1-15 was qualitatively identical). In addition to the strong membrane-associated staining, there was significant intracellular signal that was polarized toward regions of cell/cell contact. Table I summarizes the kinetic constants obtained by Lineweaver-Burke analysis of 3-0-methyl glucose uptake measurements in lines GT1-10 and GT1-15. These data are compared to previously published values obtained for GLUT-2transfected and parental AtT-SOi,, cell lines (8) or normal rat islets (16). The two GLUT-1-overexpressing cell lines exhibited a K, for 3-0-methyl glucose of 4 mM and Vmax values of 5 mmol/min/liter (GT1-10) and 6 mmol/min/liter (GT1-15).
This represents only a slight increase in K,, but a 10-fold increase in Vmar, relative to untransfected AtT-SOin, cells ( K , = 2, Vmax = 0.5). Despite these changes, the K , and Vmax values for 3-0-methyl glucose uptake into GLUT-1-expressing cells are significantly lower than the GLUT-2-expressing lines CGT-5 and CGT-6 ( K , = 17-18 and Vmax = 18-25) or normal rat islets ( K , = 18, V,, = 24). We have not measured the number of glucose transporters present in GLUT-1-uersw GLUT-2-expressing cells, so no firm conclusions about differences in intrinsic activities or turnover numbers can be made at this time. It should be noted, however, that the RNA blot hybridization analysis shown in Fig. 1 was performed with GLUT-1 and GLUT-2 probes of similar specific activities, with a clearly enhanced signal obtained for the GLUT-1 probe.
As an initial screen, we measured insulin secretion in response to glucose, forskolin, and the combination of these secretagogues in the GLUT-1-and GLUT-2-transfected lines using the previously described static incubation protocol (8).

Kinetics of 3-0-methyl glucose uptake into islets and
AtT-20,cell lines Kinetic constants derived from double-reciprocal plots of glucose uptake uersw 3-0-methyl glucose concentration in GLUT-2-transfected cell lines CGT-5 and CGT-6, GLUT-1-transfected cell lines GT1-10 and GT1-15, untransfected AtT-20i.. cells, and normal rat islets. The data for GLUT-1-transfected lines were generated in the current study. Other values are presented for comparison and are taken from Ref. 8 (*)  As shown in Table 11, all of the GLUT-2-transfected AtT-20in. lines exhibited a glucose-stimulated insulin secretion response (range of 70-260% increases relative to cells incubated in the absence of secretagogues) as well as a potent response in the presence of glucose + forskolin (range of increases of 240-680%), while all of the GLUT-1-transfected lines or cells transfected with the expression vector lacking a GLUT insert (CTC-P) responded slightly to glucose (range of 10-60%) or glucose + forskolin (17-95%), or not at all. The small response to glucose + forskolin in CTC-P or GLUT-1transfected cells is not due to a generally diminished regulated secretory pathway, since the combination of 5 FM forskolin and 1 mM isobutylmethylxanthine exerted a 4-5-fold stimulatory effect on insulin release in all three cell types (data not shown). The data for the different types of cell lines (transfected with GLUT-2, GLUT-1, or vector lacking a GLUT insert) are summarized in Fig. 3 and expressed as the percent increase in insulin release induced by glucose or glucose + forskolin relative to cells incubated in the absence of either secretagogue. Pooling of the data in this manner highlights the significant insulin secretion response to glucose ( p 5 0.02) or glucose + forskolin ( p = 0.002) in GLUT-2-transfected cells relative to GLUT-1-expressing cells. Fig. 3 also shows the insulin secretory responses over a range of glucose concentrations. The finding of maximal insulin secretion at a glucose concentration of 5100 p~ in the GLUT-2-transfected cells is consistent with our previous studies in which  and CGT-6 cells were found to be maximally responsive at 10-50 glucose (8). Perifusion experiments were carried out to study the dynamics of insulin release from the various AtT-20in. cell lines and to evaluate whether glucose-stimulated insulin secretion from GLUT-2-expressing AtT-20i, cells occurs in a time frame that resembles the rapid islet p-cell response. In the representative experiment shown in Fig. 4.4, lines CGT-6 (GLUT-2 transfected), GT1-15 (GLUT-1 transfected), and the parental AtT-20in. cells are compared. During the first 25min-perifusion period with HBSS lacking glucose, there was a gradual decline in insulin release from all three cell lines. Switching to HBSS buffer containing 5 mM glucose resulted in a 10-fold increase in insulin release from CGT-6 cells. This increase was sustained in two samples (representing a total of 5 min), after which insulin secretion declined to a second plateau that was %fold above the preglucose level. Only small changes in insulin release were observed during this period for both the parental AtT-ZOin. cells and the GLUT-l-transfected GT1-15 line. Upon removal of glucose from the perifusate, insulin secretion from the CGT-6 cells persisted at the glucose-stimulated level for approximately 10 min, but then declined rapidly. The low level of insulin release from parental AtT-20in, cells and GT1-15 cells was further reduced during perifusion with glucose-free media. In the next phase of glucose stimulation, the CGT-6 cells again showed a clear secretory response to glucose, but the response was less rapid (requiring 15 min to reach maximum), and was without an obvious first and second phase. Switching back to buffer lacking glucose again resulted in a dramatic, albeit delayed reduction in insulin release from CGT-6 cells. At the end of the experiment, cells were perifused with HBSS containing the combination of 5 mM glucose and 0.5 pM forskolin. In keeping with the static incubation data presented here and elsewhere (8), GLUT-2-expressing CGT-6 cells exhibited a much stronger insulin secretory response to glucose + forskolin than either the parental cells or the GLUT-1-transfected line. The response of line CGT-6 to glucose + forskolin was sustained until the end of the experiment, indicating that the cells were not depleted of insulin during the perifusion. Consistent with this interpretation, only slight changes in insulin content were noticed in cells (all lines) when measurements were made before and after the perifusion (data not shown).
In order to establish the consistency of the glucose response in GLUT-2 expressing cells, we performed abbreviated versions of the perifusion experiment in Fig. 4A on several of our newly isolated cell lines. Fig. 4B shows results for two GLUT-2-transfected lines, GT2-9 and GT2-12, shown with CGT-6 and vector-transfected cells as controls. Fig. 4C shows a representative perifusion of two GLUT-1-transfected lines, GT1-5 and GT1-8, again shown with a separate CGT-6 experiment and the vector-transfected control cells. GLUT-2-transfected cell lines exhibited consistently increased and rapid insulin release in response to glucose in the perifusion medium, although the magnitude of the response and the appearance of first and second phases of secretion were variable in the different CGT-6 experiments and among the different lines. GLUT-1-transfected cells in contrast consistently failed to respond to glucose perifusion.
In an attempt to probe the mechanism underlying the capacity of GLUT-2, but not GLUT-1, to confer glucose sensing, the metabolic consequences of overexpression of these transporters was investigated. We reasoned that if the differences in glucose-stimulated insulin release in the two lines were due to lower numbers of GLUT-1 transporters or lower intrinsic activity of GLUT-1, that this should be reflected in lower rates of glucose metabolism. Thus, the concentration dependence of glucose usage was measured by following production of 3H20 from [5-3H]glucose, using lines CGT-6 and GT2-12 (GLUT-2 transfected), GT1-5 and GT1-8 (GLUT-1 transfected), and untransfected cells. Measurements were performed for 5 and 15 min and carried out over a range of glucose concentrations from 10 pM to 20 mM (data shown are for 15-min experiments; similar profiles were obtained with 5-min measurements). As shown in Fig. 5, the concentration-dependent rate of glucose usage appeared similar for all of the cell lines tested. Eadie-Hofstee transformation of these primary data revealed that the "K," or for the glucose dependence of glucose metabolism averaged 1.9 mM for the GLUT-2-transfected lines, 1.6 mM for the GLUT-1 lines, and 2.5 mM in untransfected AtT-20i,, cells.
The fact that GLUT-2 transfection enables glucose-stimulated insulin release in AtT-SO,,, cells without a significant enhancement of glucose metabolism relative to untransfected or GLUT-1-transfected cells could be taken as evidence that glucose sensing in GLUT-2 expressing AtT-20i,. cells occurs via a metabolism-independent mechanism. To test this possibility, we evaluated the insulin secretory response in the GLUT-2-expressing CGT-6 cell line to two non-metabolizable analogs of glucose, 2-deoxyglucose, and 3-0-methyl glucose. As shown in Fig. 6, 2-deoxyglucose fails to stimulate insulin release, while 3-0-methyl glucose induces only a slight, statistically insignificant response. Furthermore, while D-glucose provides a strong potentiation of the forskolin-induced response in GLUT-2-expressing cells, the combination of 2deoxyglucose + forskolin produces a secretory response that is unchanged relative to forskolin alone, while the combina- tion of 3-0-methyl glucose + forskolin provides only a mild enhancement that is statistically less effective than D-glucose + forskolin ( p < 0.05). These data suggest that GLUT-2expressing AtT-BOi,. cells are similar to islet P-cells in that glucose metabolism is required for a secretory response to the sugar.

DISCUSSION
We have previously shown that overexpression of GLUT-2 in AtT-20in. cells confers glucose transport kinetics that are indistinguishable from normal rat islets. In addition, GLUT-2-expressing AtT-POin, cell lines exhibit glucose-stimulated insulin secretion, glucose potentiation of non-glucose secretagogues, and an increase in insulin content (8). The glucosestimulated insulin secretion response in engineered AtT-20in, cells was found to be maximal at 10-50 p~ glucose in our previous studies. This finding was consistent with the fact that glucokinase activity is reduced and hexokinase activity is increased in AtT-20in. cells relative to normal islets (8). Our interpretation of this data is that GLUT-2 transfection increases glucose entry into engineered AtT-POin, cells, but that in the face of dominant hexokinase activity, glucose metabolism and insulin release are maximal at subphysiological glucose concentrations. Further support for such a model is gained from experiments in which hexokinase activity has been increased by molecular manipulations in fetal islets (17) or increases naturally as a function of passage number in insulinoma cell lines (18). In both instances the dose dependence of glucose-induced insulin production or secretion is shifted to the left. Based on the model advanced above, one might predict that any glucose transporter that is capable of substantially increasing glucose entry into transfected AtT-20in, cells should serve to activate glucose-stimulated insulin release. In the current work, we have tested this hypothesis by overexpres-cfects in AtT-BOi, Cells sion of the cDNA encoding GLUT-1. This transporter is expressed in a large number of tissues, most notably brain and erythrocytes, and is 55% identical in amino acid sequence to 20). It has been shown by a number of investigators to have a lower K,,, for glucose than GLUT-2 (16, 21-23), but since its K,,, is in the low millimolar range, it should provide a capacity for glucose transport that is not rate limiting for a metabolic pathway controlled by hexokinase, an enzyme with a K, for glucose in the range of 50 p~ (24).
The strategy for the creation of stable, GLUT-l-overexpressing AtT-20in, lines was identical to that employed for GLUT-2 expression in these cells. We found that transfection with the PCB-7 plasmid in which the GLUT-1 cDNA is inserted next to the cytomegalovirus promoter resulted in as much as a 100-fold increase in GLUT-1 mRNA relative to the modest endogenous expression of this transporter in untransfected AtT-20in, cells. GLUT-1 immunofluorescence at the plasma membrane of transfected AtT-POin, cells was also dramatically increased. The increased expression of GLUT-1 translated into a small increment in the apparent K,,, for 3-0-methyl glucose from 2 to 4 mM and a more impressive 10fold increase in maximal velocity from 0.5 to 5-6 mmol/min/ liter cell space. Interestingly, the increase in maximal velocity observed in two independent GLUT-1-transfected clonal cell lines was only about one-third as great as the increase observed in two independent GLUT-2 transfected lines (Table  I). Previous studies have indicated that kinetic constants calculated for 3-0-methyl glucose uptake closely reflect the kinetics of D-glUCOSe uptake for both GLUT-1-and GLUT-2-dominated systems (23).
Glucose-stimulated insulin secretion was studied in six GLUT-1 and nine GLUT-2-transfected AtT-ZOi,. cell lines. Static incubation experiments clearly show that GLUT-2transfected cells consistently respond to glucose as an insulin secretagogue while GLUT-1-transfected cells have only a very modest response. In keeping with our previous work (8), glucose-stimulated insulin secretion was found to be maximal a t glucose concentrations in the range of 50 pM in GLUT-2transfected cells. Furthermore, administration of the combination of glucose + forskolin resulted in a dramatically enhanced insulin secretory response in GLUT-2-transfected cells.
Perifusion experiments were performed in an effort to confirm these results and to simultaneously learn more about the dynamics of insulin release. Strikingly, GLUT-2-transfected cells, but not GLUT-1-transfected or vector-transfected cells, dramatically increased their insulin output within minutes of a change from HBSS buffer lacking glucose to the same buffer containing 5 mM glucose. Furthermore, return to HBSS lacking glucose resulted in reduction of insulin release from the GLUT-2-expressing cells. The finding of generally consistent effects of GLUT-2 uers'sus GLUT-1 transfection in a large number of independent clonal lines in this study strongly supports the idea that the effects observed are a consequence of the expression of the particular glucose transporter isoform, and are not due to clonal variability.
We have considered three models that might potentially explain the divergent effects of GLUT-1 and GLUT-2 expression in AtT-BOin, cells, as summarized in Fig. 7. In the first of these, we assume that the number of functional GLUT-2 molecules exceeds the number of functional GLUT-1 molecules at the plasma membrane surface for the respective transfection experiments. The second, related model would hold that the number of glucose transporter molecules is similar for the two transfection experiments, but that GLUT-

Glucose Transporter Effects in
AtT-20im Cells 15211 2 has a higher intrinsic specific activity than GLUT-1. For either of these models, the specific effects of GLUT-2 in mediating glucose-stimulated insulin release would be explained by enhanced glucose entry.

Glutl
Direct measurement of glucose uptake in the transfected lines revealed that maximal velocity of glucose uptake in GLUT-1-transfected cells was indeed only one-third of that in GLUT-2-transfected cells. Despite the large differences in glucose uptake capacity among cell lines, the rate of glucose usage was found to be similarly affected by the glucose concentration in GLUT-2-and GLUT-1-transfected or control cells, with an So.5 for glucose metabolism in the range of 2 mM in all cases. These observations suggest that glucose uptake is not rate limiting for glucose metabolism in AtT-20i,. cells, and that the rate of metabolism is instead limited by a low K,,, component (possibly hexokinase). Taken together, the foregoing observations seem to provide support for the third of the proposed mechanisms in Fig. 7, namely that GLUT-2, but not GLUT-1, is capable of forming productive interactions with other components of insulin-secreting cells, and that this putative interaction is critical for functional glucose signaling. Potential partners for GLUT-2 coupling might include signal transduction molecules such as GTP-binding proteins. Alternatively, GLUT-2 might serve to nucleate complexes of metabolic proteins or enzymes capable of substrate "channeling" in a manner analagous to complexes or "metabolons" that occur among enzymes of the citric acid cycle (25).
While there is no direct evidence for participation of GLUT-2 in metabolic channeling of glucose, other reactions of the glycolytic sequence may be tightly coupled. Recently, Malaisse and Bodur (26) have postulated that coupling occurs within the enzyme sequence hexokinase/glucokinase-phosphoglucoisomerase-phosphofructokinase in order to account for generation of 3H20 from [2-3H]glucose in intact islets at a rate lower than that predicted by activities of these enzymes measured in uitro. It has also been proposed that the islet isoform of glucokinase, which is different from liver glucokinase in that it contains a highly charged N terminus (2, 7, 27), might be equipped for physical coupling with the GLUT-2 transporter and more efficient glucose utilization (2). The presence of GLUT-2 may promote the formation of an enzyme complex, such as the one described by Malaisse, in close proximity to the plasma membrane, enhancing glycolytic flux within this compartment and leading to changes in membrane potential. Such a model has the added benefit of potentially explaining a long-standing paradox of glucose signaling in pcells, which is that ATP-sensitive K' channels become inhibited upon exposure of islet cells to glucose (28, 29), despite a very limited alteration in whole cell ATP/ADP ratio (30). Localized metabolism of glucose mediated by a GLUT-2 nucleated metabolon might allow effective local production of ATP in close proximity to adjacent K+ channels. Precedent for such a model has in fact been advanced for myocytes, in which production of ATP in response to administration of glycolytic substrates has been shown to be more effective in closing K+ channels than ATP produced by addition of Krebs cycle substrates (31).
We must also consider the possibility that the specific capacity for glucose signaling conferred by GLUT-2 transfection is only relevant to the non-islet AtT-20in, cell lines, and that glucose signaling in such cells occurs by a mechanism distinct from that operative in islet-derived cells. Although fuel-mediated signaling pathways of AtT-BOi,, cells may ultimately prove to differ from islet-derived cells in some aspects, a fundamental similarity of the two cell types is the apparent requirement for glucose metabolism for stimulated insulin release, as illustrated by the absence of a direct or potentiating response to non-metabolizable analogs of glucose. Since the overall rate of glucose metabolism is similar in GLUT-1-and GLUT-2-transfected cells, the glucose analog data suggest that either a portion of glucose flux is localized or channeled in GLUT-2-transfected cells, or alternatively, that the GLUT-2 protein provides a necessary but not sufficient signal that requires glucose metabolism for transmission. It should also be noted that the time frame of the response in perifusion experiments is extremely rapid, indicating that glucose is generating a specific signal, as opposed to having a generalized effect on cell viability. Finally, our recent studies show that transfection of glucose-unresponsive rat insulinoma cells greatly enhances their capacity for glucose-stimulated insulin release.' In sum, the AtT-BOi,, cell system may ultimately prove to be relevant to understanding of the glucose-sensing pathway in normal islet cells.
It is of interest to compare the results reported herein, obtained by engineered expression of glucose transporter isoforms in insulin-secreting cells with other recent studies in which various experimental factors have been used to alter the relative levels of expression of these transporters. In agreement with our data, Miyazaki et al. (32) isolated a number of clonal cell lines from transgenic animals in which T-antigen expression directed by the insulin promoter was used to cause P-cell tumors. The MIN-6 cell line expressed GLUT-2 abundantly and exhibited a significant glucose-stimulated insulin secretion response. In contrast, the MIN-7 line, isolated in parallel with MIN-6, was found to express GLUT-1 predominantly and very little GLUT-2; these cells failed to respond to glucose (32).
A somewhat different picture emerges from studies with isolated islets. It has recently become appreciated that transfer of islolated islets to a tissue-culture environment causes increased expression of GLUT-1 and some reduction, but not a loss of GLUT-2 expression; this trend is most prominent when the islets are cultured at low glucose concentration (3 mM glucose) (33, 34). Interestingly, partial loss of GLUT-2 expression is correlated with complete absence of glucosestimulated insulin release not only in isolated islets cultured at low glucose, but also in animal models of non-insulindependent diabetes mellitus (4). The partial loss of immunocytochemically detectable GLUT-2 that occurs upon transfer of islets to tissue culture can be largely overcome by culturing the islets a t high glucose concentrations (30 mM) for 7 days (34). Culturing of islets a t high glucose also causes a return of the glucose-stimulated insulin secretion response. Despite the return of immunologically detectable GLUT-2 after 7 days of culture at high glucose, glucose uptake was found to be decreased relative to freshly isolated islets. These data were used by Tal et al. (34) to argue that regulation of GLUT-2 activity is secondary to other factors, in particular the regulation of glucokinase enzymatic activity in the control of glucose-stimulated insulin release. This argument is based on the knowledge that the level of glucokinase activity is closely correlated with the magnitude of the insulin secretory response in islets cultured at low and high glucose concentrations (35, 36). While aspects of this formulation are likely to be correct, the data are also consistent with the new model advanced herein, in which GLUT-2 protein is required for effective signal transduction through physical coupling with other islet components. The loss of glucose-stimulated insulin release that occurs with only a partial loss of GLUT-2 protein in both the in vivo and in vitro models may fit with our model if a minimum level of GLUT-2 expression is required for effective interactive signaling, or alternatively, if loss of GLUT-2 expression is specific for a subpopulation of @-cells that are particularly competent for glucose sensing. Indeed, such functional heterogenity in 0-cells is now well accepted (37,38) and may fit with recent observations suggesting that glucokinase is expressed in only a subset of P-cells (39).
In conclusion, elucidation of the mechanism for glucosestimulated insulin release in the AtT-20i,, cell model will require more extensive studies of the effect of GLUT-2 expression on glucose metabolism and signal transduction in these cells. Specific approaches that may be informative include structure-function studies of GLUT-2 and modulation of the hexokinase/glucokinase ratio within the model system described in this study.